Skip to content

🔭 milk

High-performance real-time image processing framework for Adaptive Optics and scientific computing. Microsecond-latency pipelines through zero-copy shared memory.

GitHub stars License Build

milk orchestrates many small compute units that communicate through zero-copy shared memory tensors, enabling microsecond-latency data pipelines. The three pillars — ImageStreamIO, FPS, and processinfo — live entirely in /dev/shm/.

🚀 Getting Started

  • Install

    Clone, build, and configure the milk framework.

    Installation

  • Build Tiers

    Engine → Core → Full: compile only what you need.

    Build tiers

  • CLI Overview

    Interactive shell, standalone executables, and scripting basics.

    CLI overview

  • FAQ

    Common issues with builds, SHM, FPS, and CLI.

    FAQ & Troubleshooting

🏛 Core Concepts

  • Streams

    Zero-copy shared memory tensors (ImageStreamIO).

    Streams

  • FPS

    Live parameter sync, state control, TUI dashboards.

    FPS

  • Process Info

    Heartbeat telemetry, loop-rate profiling, health monitoring.

    Process Info

  • Architecture

    System overview, layered design, data flow diagrams.

    Architecture

🛠 Developer Guides

  • Tutorial

    Write your first compute module step by step.

    Tutorial

  • Coding Standards

    C style, line length, includes, Kernel-Doc.

    Coding standards

  • Adding Plugins

    Build modules that compile alongside the core.

    Plugins

  • Template Code

    Breakdown of milk_module_example.

    Template source

📊 Operations & Reference

  • Performance

    CPU pinning, RT scheduling, SIMD, BLAS, GPU.

    Performance

  • PGO & LTO

    Profile-guided optimization + static link-time optimization for 15–40 % speedup.

    PGO & LTO

  • Debugging

    GDB, tmux logs, procinfo diagnostics, common failures.

    Debugging

  • API Reference

    Auto-generated Doxygen C API docs with call graphs.

    Doxygen API

🔗 More Resources